70 likes | 239 Views
NFSv4 Migration Document Status. David Noveck Nfsv4 Working Group meeting at IETF88 November 5, 2013. To be Discussed. draft-ietf-nfsv4-rfc3530-migration-update draft-ietf-nfsv4-rfc5661-migration-update draft-ietf-nfsv4-migration-issues Migration implementation work
E N D
NFSv4 Migration Document Status David Noveck Nfsv4 Working Group meeting at IETF88 November 5, 2013
To be Discussed • draft-ietf-nfsv4-rfc3530-migration-update • draft-ietf-nfsv4-rfc5661-migration-update • draft-ietf-nfsv4-migration-issues • Migration implementation work • For NFSv4.0, considerable work done • Chuck Lever’s presentation will cover • NFSv4.1 just starting
draft-ietf-nfsv4-rfc3530-migration-update-03 • Addressed issue of NFS4ERR_DELAY and RELEASE_LOCKOWNER • Return allowed for migration • Restrictions to address client compatibility • RECOMMENDATION for FS-grace if no TSM • If neither, lock loss problem • Updated discussion of clientid-related errors • Reworked discussion of owner-based info • Should reduce need for testing individual stateids
draft-ietf-nfsv4-rfc3530-migration-updateFuture Revisions • Needed for future revisions • Clarifications • Other changes prompted by wg discussion • Issues which arise from implementation experience • Change so we update rfc3530bis RFC (and not 3530) • Probably need to wait for rfc3530bis approval • Looking for WG last call around IETF89
draft-ietf-nfsv4-rfc5661-migration-update • Work on draft starting now • Needed since implementation work is starting. • Want to hear about people’s needs in this area • Comparison with v4.0 document: • Client-id string approaches (no need in v4.1) • Lease-merger needed also • Need to address pNFS-related issues • Co-ordination with locations_info stuff • Other v4.1-related areas people are interested in
Stateid issues for v4.1 migration • V4.1 stateids are qualified by clientid • Effective stateid space is bigger than in v4.0 • Trying to separately assign stateids of a per-fs basis • Tempting but it runs into trouble • Problem with FREE_STATEID and TEST_STATEID • No way to specify the fh • Hard to change at this point, even if, in retrospect, it’s a mistake • Better extension mechanism would help.
draft-ietf-nfsv4-migration-issues-04 • Explanation of issues and choices in this area • Covers both minor versions • Pretty complete coverage of v4.0 • Treatment of v4.1 is more limited right now • Recent change in -04 • Mention option for SETCLIENTID_PLUS • Document future • Will update to reflect any issues found in doing rfc5661 migration update • May eventually publish as informational RFC